Administrative data and inverse occurrence of cancer and Alzheimer's disease in elderly people: a methodological approach to control for biases due to under diagnosis.

Introduction Lower incidence of cancer in persons with Alzheimer disease and vice-versa has been previously reported, however this might be due to an under-diagnosis of the second disease after the occurrence of the first. Aim of this study was to estimate the unbiased co-occurrence of the two diseases in a population of about one million persons, using administrative health data (AHD). Materials and Methods In AHD we identified incident Cancer and AD cases over a period of five years. To control for under-diagnosis of the second disease three analyses were carried-out: for times preceding or following the index diagnosis, for surviving or not the follow-up period and for different types and sites of cancer. The relative risks (RRs) of AD in persons with cancer and vice-versa, relative to general population, were estimated as observed timed expected cases. Results The RRs of cancer in persons with AD and of AD in persons with cancer were both reduced. The inverse relationship of occurrence was independent from the temporal order of appearance of the diseases, from life-expectancy reduction and from different cancer types. Discussion Administrative data represent a valid source for the conduction of cohort studies on the co-occurrence of diseases.